R E G U L A R P A P E R
Adaptive hybrid error correction model for video streaming
over wireless networks
Ming-Fong Tsai•Tzu-Chi Huang•Chih-Heng Ke•
Ce-Kuen Shieh•Wen-Shyang Hwang
Published online: 9 November 2010 Ó Springer-Verlag 2010
Abstract The Hybrid ARQ (HARQ) mechanism is the well-known error packet recovery solution composed of the Automation Repeat reQuest (ARQ) mechanism and the Forward Error Correction (FEC) mechanism. However, the HARQ mechanism neither retransmits the packet to the receiver in time when the packet cannot be recovered by the FEC scheme nor dynamically adjusts the number of FEC redundant packets according to network conditions. In this paper, the Adaptive Hybrid Error Correction Model (AHECM) is proposed to improve the HARQ mechanism. The AHECM can limit the packet retransmission delay to the most tolerable end-to-end delay. Besides, the AHECM can find the appropriate FEC parameter to avoid network
congestion and reduce the number of FEC redundant packets by predicting the effective packet loss rate. Meanwhile, when the end-to-end delay requirement can be met, the AHECM will only retransmit the necessary number of redundant FEC packets to receiver in compari-son with legacy HARQ mechanisms. Furthermore, the AHECM can use an Unequal Error Protection to protect important multimedia frames against channel errors of wireless networks. Besides, the AHECM uses the Markov model to estimate the burst bit error condition over wireless networks. The AHECM is evaluated by several metrics such as the effective packet loss rate, the error recovery efficiency, the decodable frame rate, and the peak signal to noise ratio to verify the efficiency in delivering video streaming over wireless networks.
Keywords Adaptive hybrid error correction model Wireless network Video streaming
1 Introduction
Wireless networks have been extensively used to construct a communication infrastructure at various places such as home, school, and airport, because of the high flexibility and low cost. Due to the emergency of multimedia information and various electronic devices capable of networking, wireless networks gradually take the responsibility to transport data for bandwidth-killer applications, e.g., video applications. However, wireless networks have the well-known high Bit Error Rate (BER) and fluctuation in the channel quality [1,2] both of which are harmful to video communication. Once bit errors occur in communication, a.k.a. channel error, wireless networks propagate the bit errors blindly without any further processing. Wireless
M.-F. Tsai C.-K. Shieh
Department of Electrical Engineering, Institute of Computer Communication Engineering, National Cheng Kung University, Tainan, Taiwan
e-mail: fone@hpds.ee.ncku.edu.tw C.-K. Shieh
e-mail: shieh@hpds.ee.ncku.edu.tw T.-C. Huang
Department of Electronic Engineering,
Lunghwa University of Science and Technology, Taoyuan, Taiwan
e-mail: tzuchi.phd@gmail.com C.-H. Ke (&)
Department of Computer Science and Information Engineering, National Quemoy University, Kinmen, Taiwan
e-mail: smallko@gmail.com W.-S. Hwang
Department of Electrical Engineering,
National Kaohsiung University of Applied Sciences, Kaohsiung, Taiwan
e-mail: wshwang@mail.ee.kuas.edu.tw DOI 10.1007/s00530-010-0213-x
networks can severely degrade video streaming quality when the receiver tries to reconstruct the video frame from the erroneous data. Accordingly, wireless networks make important the task of protecting video streaming from channel errors.
Packet loss due to channel errors is called the erasure error and traditionally recovered by either the Automation Repeat reQuest (ARQ) mechanism or the Forward Error Correction (FEC) mechanism. The ARQ mechanism uses parity check codes [3] and retransmits lost packets to the receiver when the sender fails to receive an acknowl-edgement from the receiver or receives a retransmission request from the receiver. According to channel conditions, the ARQ mechanism can protect video streaming from channel errors with the packet retransmission [4]. In con-trast to the ARQ mechanism, the FEC mechanism is more suitable to protect video streaming from channel errors with a short transmission delay and a transmission reli-ability improvement. Based on transmitting a certain number of FEC redundant packets to the receiver (i.e., the FEC redundancy), the FEC mechanism transmits n packets composed of k source packets and n - k redundant pack-ets. If the receiver receives more than or equal to k packets among the n packets, the FEC mechanism can make the receiver to recover the lost or erroneous packets from the FEC redundant packets [5,6].
The Hybrid ARQ (HARQ) mechanism [7,8] combines the ARQ mechanism and the FEC mechanism. The HARQ mechanism can use different retransmission schemes geared toward video streaming [9, 10] and various pow-erful FEC erasure coding techniques such as Reed-Solo-mon (RS) codes [11, 12]. Accordingly, The HARQ mechanism not only provides good reliability and flexi-bility as the ARQ mechanism but also achieves throughput similar to the FEC mechanism. Although many HARQ mechanisms have been proposed [13, 14] with the enhancement of different retransmission schemes [15] and the improvement of various FEC mechanisms, they have two drawbacks in common as follows. First, the HARQ mechanisms cannot retransmit the packet to the receiver in time when the packet cannot be recovered from the FEC. Second, the HARQ mechanisms cannot dynamically adjust the FEC redundancy according to network conditions. Accordingly, the present HARQ mechanisms are not enough to protect video streaming from channel errors in wireless networks.
In this paper, the Adaptive Hybrid Error Correction Model (AHECM) is proposed to improve the HARQ mechanism. The AHECM can find the maximum retrans-mission threshold to retransmit lost packets to the receiver in time at the premise of reducing the FEC redundancy. To transmit lost packets to the receiver in time, the AHECM calculates the most tolerable end-to-end delay. To reduce
the FEC redundancy, the AHECM calculates the appro-priate FEC parameter. Accordingly, the AHECM not only can avoid network congestion stemming from unlimited packet retransmission but also can reduce the FEC redun-dancy to conserve network bandwidth by predicting the effective packet loss rate. Meanwhile, when the end-to-end delay requirement can be met, the AHECM will only retransmit the necessary number of redundant FEC packets to receiver in comparison with legacy HARQ mechanisms. Most importantly, the AHECM can use an Unequal Error Protection to decide the appropriate FEC redundancy for different video frame types in order to produce the optimal amount of FEC redundancy. The AHECM uses the Markov model to approach the burst bit error condition over wireless networks. In experiments, the AHECM is evalu-ated by the effective packet loss rate, the Error Recovery Efficiency (ERE) model, the revised Decodable Frame Rate (DFR) model [16,17], and the Peak Signal to Noise Ratio (PSNR). Compared with two latest HARQ proposals, the AHECM shows the promising results when delivering video streaming over wireless networks.
We organize the remainder of this paper as follows: We review the related works in Sect. 2. Then we present the AHECM in Sect.3. After that, we introduce the definition of the ERE model and how we revise the DFR model in Sect.4. The experimental results are also shown in the same section. Finally, we conclude this paper in Sect. 5.
2 Related works
The present HARQ mechanisms can be divided into Type I HARQ scheme [11, 12] and Type II HARQ scheme [13, 14]. Type I HARQ scheme contains additional symbols for error detection and correction in transmission blocks. Type I HARQ scheme repairs and accepts a block in the receiver as the correct block, if the corrupted symbol count is lower or equal to the maximum count of repairable errors t. Conversely, Type I HARQ scheme rejects a block received by the receiver and retransmits the block, if the corrupted symbol count is higher than the maximum count of repairable errors t. Type I HARQ scheme can use the properly chosen error correcting codes to significantly reduce packet retransmission times in comparison to the ARQ mechanism. However, Type I HARQ scheme may discard non-repairable blocks in the decoder of the receiver even though they still contain some useful information.
For improving Type I HARQ scheme, Type II HARQ scheme does not discard corrupted blocks but leaves them for further processing. Type II HARQ scheme merely retransmits the stronger protection part of the previous corrupted block instead of the whole block. Type II HARQ scheme uses the decoder in the receiver to repair the stored
corrupted block according the protection symbols. Hence, Type II HARQ scheme can preserve system resources such as network bandwidth and CPU time. No matter which HARQ scheme type is used, the present HARQ mecha-nisms cannot retransmit packets to the receiver in time; accordingly, video applications on the receiver cannot utilize data carried by packets arriving late. Moreover, the present HARQ mechanisms cannot dynamically adjust the FEC redundancy according to network conditions; accordingly, the inappropriate FEC redundancy may con-gest the network.
Packet loss can happen when networks are congested with the inappropriate FEC redundancy or constructed by unstable channels. Packet loss in a hierarchical coding structure such as MPEG can substantially reduce the quality of the received video at the receiver. Various error-resilient video coding techniques have been proposed [14,15] to find the appropriate FEC redundancy. Moid and Fapojuwo [18] and Subramanian et al. [19] use a heuristic algorithm to find the optimal FEC redundancy, but do not retransmit the packet to the receiver in time when the packet cannot be recovered by the FEC mechanism. Kotuliakova and Polec [20] analyzes the throughput per-formance of pure and HARQ techniques, but neither cor-rectly finds the effective packet loss rate nor dynamically adjusts the FEC redundancy according to network condi-tions. Tan and Herfet [21] tries to choose a best scheme to minimize the FEC redundancy, but fails to get the optimal throughput performance according to its equation.
The Network-Adaptive Selection of Transport Error (NASTE) mechanism [22] uses a network-adaptive selec-tion of transport error mechanism to dynamically select one mode from the static FEC mode, the ARQ mode, the adaptive FEC and ARQ mode, and the interleaving FEC and ARQ mode. If link quality is better and packet loss rate is higher, the NASTE uses static FEC or ARQ modes. If packet loss happens sparsely, the NASTE selects the ARQ mode for efficient bandwidth usage. If packet loss happens diversely, the NASTE uses the static FEC mode instead of the ARQ mode. The NASTE recovers packet losses and uses bandwidth efficiently to control the FEC redundancy. When a packet burst happens to a wireless channel, the NASTE uses the interleaving FEC and ARQ mode. According to the FEC block size, the NASTE changes the interleaving depth in order to protect burst packet loss. However, the NASTE neither works efficiently due to discarding usable source packets in an unrecoverable FEC block nor controls the FEC redundancy appropriately when the network bandwidth is not enough.
The cross-layer error control mechanism [6] uses a cross-layer error control framework for robust and low delay multimedia streaming in tandem-connected IEEE 802.11 wireless LANs and the Internet. It models the
end-to-end delay and packet loss rate as a function of the ARQ and FEC error control mechanism employed at the appli-cation and wireless link layers. It uses an analytical model as the basis of a delay-constrained error control algorithm to adapt the protection level at the application and link layers in order to minimize the end-to-end packet loss rate. It can validate the efficiency of the proposed cross-layer error control methodology for delay-sensitive pre-com-pressed video streaming. However, it fails to work effi-ciently due to discarding usable source packets in an unrecoverable FEC block. Accordingly, proposals [6,22] discard usable source packets in an unrecoverable FEC block will have a high effective packet loss rate that is harmful to the recovery of a video frame.
Among the various error-resilient video coding tech-niques, the Unequal Error Protection (UEP) works effi-ciently on video streaming. The UEP identifies the unequal importance of video data in the hierarchical video coding structure and gives the video data different amounts of protection according to the importance. In MPEG, the UEP exploits the difference in importance among frames (i.e., the I, P, and B frames) in a GOP and allocates more protection bits to the I frame than to the P and B frames [23, 24]. Different UEP mechanisms have been proposed for various types of scalable video coding. In [25], for example, the UEP FEC is proposed to protect packet losses by using the unequal importance of packets in different layers. In [26], for another example, the UEP is applied to MPEG-4 fine grain scalable compressed video data by using rate-distor-tion informarate-distor-tion for each layer. However, the present UEP mechanisms neither use an analytical model to illustrate the performance, nor perform analyses to decide the appropriate FEC redundancy for different video frame types. Further-more, the present UEP mechanisms do not dynamically adjust the FEC redundancy according to network conditions.
3 Adaptive hybrid error correction model 3.1 AHECM overview
The AHECM collects meta-information about the average packet loss rate, the average Round-Trip Time (RTT), and the available bandwidth at the receiver. When the meta-information is changed according to channel or network conditions, the AHECM makes the receiver send the meta-information to the sender in order to adjust parameters of the AHECM. With parameter updates, the AHECM at the sender can decide the appropriate FEC redundancy for different video frame types when the sender transmits video streaming to the receiver over wireless networks. With information about the average RTT and the most tolerable end-to-end delay provided by the receiver, the
AHECM at the sender can find the maximum retransmis-sion time and retransmits the lost packet to the receiver in time at the premise of reducing the FEC redundancy. With information about the average packet loss rate and the available bandwidth provided by the receiver, the AHECM at the sender can calculate the appropriate FEC parameter to avoid network congestion and the unnecessary FEC redundancy. Meanwhile, when the end-to-end delay requirement can be met, the AHECM will only retransmit the necessary number of redundant FEC packets to receiver in comparison with legacy HARQ mechanisms. The AHECM uses a Markov model to accurately predict the effective packet loss rate at the premise of considering the burst bit error condition in wireless networks. The AHECM uses an UEP algorithm to give various frames different protections. Accordingly, the AHECM is the critical component of our proposal in this paper.
3.2 AHECM philosophy
We assume that the video streaming needs transmission bandwidth VSRB bit per second. We also use the Packet
Gap Model (PGM) [2] to measure the available bandwidth in this paper. The PGM exploits the information in the time gap between the arrivals of two successive probes at the receiver. A probe pair is sent with a time gap Din; it reaches
the receiver with a time gap Dout. Assuming a network
bottleneck exists along the path from the sender to the destination and the queue in the bottleneck does not become empty between the departure of the first probe in the pair and the arrival of the second probe, then Doutis the
time taken by the bottleneck to transmit the second probe in the pair and the cross traffic that arrived during Din.
Thus, the time to transmit the cross traffic is (Dout - Din),
and the rate of the cross traffic can be calculated as (1). CDout Din
Din
ð1Þ where C is the capacity of the bottleneck and the unit is bit per second. Moreover, the Available Bandwidth (AB) (bit per second) can be calculated as shown in (2):
AB¼ C 1 Dout Din Din
: ð2Þ
For deducing the AB to get the appropriate FEC redundancy without committing congestion loss, we can calculate the maximum FEC coding rate at the sender according to (3). In the initial case, the ABinitialis for video
streaming and FEC redundancy. During the transmission, the ABtransmissionwill be smaller than ABinitial because the
ongoing video streaming and FEC packets have already consumed some bandwidth. Hence, the FEC redundancy during the transmission case needs recalculation:
FECcoding rate
¼
ABinitialVSRB
ABinitial
; initial
ABtransmissionþVSRBFECcoding rate previous
ABtransmissionþVSRB 1þð FECcoding rate previousÞ; transmission
8 < :
:
ð3Þ The FEC mechanism uses an (n, k) block erasure code to convert k source packets into a group of n coded packets. The end-to-end delay of AHCEM successfully receives one FEC block including propagation delay time, transmit delay time, and round-trip time to support packet retransmission.
If the most tolerable end-to-end delay provided by the receiver is TMAX with delay constraints [27,28]
custom-ized by users or applications, we can calculate the FEC block end-to-end delay in the retransmission as shown in (4):
TOWDþ n Tð INTþ TTRAÞ þ R TRTTþ Nrðn; kÞ
Tð INTþ TTRAÞ TMAX ð4Þ
where TOWDis the one-way delay, TRTT is the round-trip
time, TINTis the transmission gap, TTRAis the transmission
time of one packet, n is the FEC block size, R is the number of retransmissions, and Nr is the total number of
retransmission packets. When the value of R is 1, we can get the total number of retransmitted packets Nr_one
according to (5). The number of retransmitted packets is the result that the number of packets in one FEC block subtracts the number of packets which successfully received in unrecovered FEC blocks.
Nr one¼ n
Xk1 i¼0
Cin ð1 pÞi pni i ð5Þ
where p is the average packet loss rate reported by the receiver at run time. When the value of R is 2, we know that Nris equal to Nr_one? Nr_two. We can use (6) to get
Nr_two, the number of retransmitted packets, which is the
result that the number of packets in one FEC block subtract the number of packets which successfully received in unrecovered FEC blocks and the number of packets which successfully received in unrecovered FEC blocks after the first packet retransmission.
Nr two¼ n Xk1 i¼0 Cin 1 pð Þipni i þ X ki1 j¼0 Cnij 1 pð Þjpnij j ! ð6Þ
According to Eqs. (3) and (4), we can get the candidates of suitable FEC parameters that meet the maximum retransmission threshold without incurring network congestion. According to (3), we can get the FEC coding
rate for transmitting video frames without incurring network congestion. With the maximum retransmission threshold in (4), we can make sure that the video streaming can be delivered to the receiver in time. For protecting burst packet losses on demand, we can use an interleaving equation [22], i.e., (7), which is transformed from (4). TOWDþ n Tð INTþ TTRAÞ þ R TRTTþ Nrðn; kÞ
Tð INTþ TTRAÞ
TMAX
D ð7Þ
where D is the block interleaving depth. Finally, we can configure the appropriate value of n, k, and R in (7) in order to get the effective packet loss rate.
We address the UEP algorithm used by the AHECM. We control the FEC coding rate at the sender according to (3) in order to avoid network congestion. We give impor-tant frames (i.e., I frame) a better protection with more FEC redundancy by means of a high FEC coding rate at the sender. After processing important frames, we give unim-portant frames (e.g., P frame or B frame) the residual FEC redundancy. We assume that a GOP of a video streaming file has NIpackets for I frame, NPpackets for P frame, and
NB packets for B frame. We use N in (8) to denote all
packets of the GOP belonging to the video streaming file.
N¼ NIþ NPþ NB ð8Þ
We assume that each I frame comprises kIvideo source
packets and hIFEC redundant packets, and nIis number of
kI? hIpackets, so a total of approximate NI/kIblocks can
be transmitted over networks for the video streaming file. With the average packet loss rate Ppkt reported by the
receiver, we can use (9) to get the probability of a block that waits for being retransmitted for recovery.
PIblock ¼ X kI1 i¼0 CnI i 1 Ppkt i Ppkt nIi ð9Þ
Moreover, we can use (10) to get the probability of a block that has been retransmitted one time for recovery.
PIblock ¼ X kI1 i¼0 CnI i 1 P pkt i P pkt nIi X kIi1 j¼0 CnIi j 1 Ppkt j Ppkt nIij ð10Þ
The receiver can receive (k - 1) or fewer source data packets under the unsuccessfully recovered FEC blocks. Those source data packets also can transmit to the application layer for video streaming application decoding. Hence, the effective packet loss rate is gotten by subtracting the remaining source data packets from the number of packets in the unsuccessful recovered FEC blocks.
Hence, we can estimate the effective packet loss rate used by the FEC decoding process at the receiver according to (11).
PIeffective¼ PIblock
1 kIþ hI
NrðkIþ hI; kIÞ ð11Þ
We assume that each P frame in the transmission comprises kP source packets and hP FEC redundant
packets, and nP is number of kP? hP packets, so a total
of approximately NP/kP blocks can be transmitted over
networks. Similarly, we can use (12) to estimate the effective packet loss rate used by the FEC decoding process at the receiver.
PPeffective¼ PPblock
1 kPþ hP
NrðkPþ hP; kPÞ ð12Þ
We assume that each B frame in the transmission comprises kB source packets, hB FEC redundant packets,
and nB is number of kB? hB packets to make a total of
approximately NB/kB blocks delivered over networks.
Similarly, we can use (13) to estimate the effective packet loss rate used by the FEC decoding process at the receiver.
PBeffective¼ PBblock
1 kBþ hB
NrðkBþ hB; kBÞ ð13Þ
We can use (14) to estimate the total effective packet loss rate used by the FEC decoding process at the receiver.
Peffectivepkt
¼PIeffective NIþ PPeffective NPþ PBeffective NB
NIþ NPþ NB
ð14Þ Considering successful and unsuccessful deliveries of I frame, P frame, and B frame, accordingly, we can use Eqs. (15), (16), and (17) to estimate the recovery overhead. In the recovery overhead equations, the FEC redundancy and retransmission packets are recovery overhead when the FEC block can be recovered. The whole FEC block and retransmission packets are recovery overhead when the FEC block cannot be recovered.
OverheadI¼ 1 Pð IeffectiveÞ hð Iþ NrðnI; kIÞÞ þ PIeffective nð Iþ NrðnI; kIÞÞ ð15Þ OverheadP¼ 1 Pð PeffectiveÞ hð Pþ NrðnP; kPÞÞ þ PPeffective nð Pþ NrðnP; kPÞÞ ð16Þ OverheadB¼ 1 Pð BeffectiveÞ hð Bþ NrðnB; kBÞÞ þ PBeffective nð Bþ NrðnB; kBÞÞ ð17Þ
In other words, we can use (18) to estimate the total recovery overhead for transmitting a GOP of a video streaming file over wireless networks.
Overhead¼ OverheadIþ OverheadPþ OverheadB ð18Þ
Finally, we can use (18) at the receiver to deduce the minimum recovery overhead, i.e., the appropriate FEC parameter. To this end, we use (19) to find the best pair (n, k, R) from the candidates produced from Eqs. (3) and (4).
n; k; R
ð Þ ¼ arg min Overheadð Þ ð19Þ
With the best pair (n, k, R), we can decide the optimal amount of FEC redundant packets (i.e., the optimal FEC redundancy) transmitted to the receiver over wireless networks. Because R in the best pair (n, k, R) denotes the maximum retransmission threshold, we can make the AHECM retransmit lost packets to the receiver in time at the premise of reducing the FEC redundancy and avoiding network congestion.
3.3 Other concerns of AHECM
As a concrete model, the AHECM is concerned with burst bit errors because burst bit errors often happen in wireless networks. If the bit error rate (BER) and burst bit error length can be known at advance for wireless networks in the burst bit error conditions, the AHECM can use a Markov model [5] to accurately predict the effective packet loss rate. The AHECM uses the Markov model in Fig.1 where the G state presents the successful transmission and the B state presents the unsuccessful transmission. We use Pbg to express the probability that a bit in a transmission
stays in the B state but enters the G state in the next transmission. We can calculate Pbgaccording to (20).
Pbg¼
1 Lburst
ð20Þ where Lburst is the average burst bit error length.
Accordingly, we know that a bit has a small probability to enter the G state when the burst bit error length is long. Conversely, we use Pgbto express the probability that a bit
in a transmission stays in the G state but enters the B state in the next transmission. We can calculate Pgbaccording to
(21). Pgb¼ Pbg
BER
1 BER ð21Þ
Accordingly, we can use (22) to get the probability of the successful transmission for a packet in wireless net-works having different BERs and burst bit error lengths. p¼ 1 BERð Þ 1 P gb
8MTU1
ð22Þ where MTU is the maximum transmission unit [5]. Note that the first bit of a packet should be in the G state and the remaining bits of a packet should have (1 - Pgb)
proba-bility to stay in the G state.
4 Experimental results
First of all, we use the effective packet loss rate, the well-known matrix, to evaluate the AHECM and compare the AHECM with two latest HARQ mechanisms. Then, we propose the Error Recovery Efficiency (ERE) model to observe the cost of packet recovery in the three mecha-nisms. Moreover, we revise the Decodable Frame Rate (DFR) model [16, 17] to observe the number of frames recovered by the receiver because the original DFR model merely observes the average packet loss rate of all frames rather than that of each frame. We explain the ERE model and the way to revise the DFR model before beginning the effective packet loss rate experiment, the ERE experiment, and the DFR experiment. To observe the impacts on the video quality at the receiver side, we, respectively, present the PSNR values (objective) and several visual video (subjective) for these different three mechanisms.
4.1 Error recovery efficiency model
Normally, the sender can make an effective improvement in the recovery performance at the receiver by sending a great number of FEC redundant packets. However, the sender may overload the network unnecessarily with many redundant packets when the packet loss rate is small. Accordingly, the sender needs a performance model to evaluate the utilization of the redundant packets in the recovery process. We propose the Error Recovery Effi-ciency (ERE) model as a recovery effiEffi-ciency matrix. We define the ERE model by the ratio of the total number of recovered packets with a HARQ mechanism to the total number of redundant packets sent during the course of the transmission. The unit of ERE model is percentage and we use (23) to express the ERE model. We say that the recovery efficiency is good if the ERE value is high. ERE¼ N p Peffectivepkt
Total number of redundant packets ð23Þ 4.2 Revision of decodable frame rate model
According to the hierarchical structure of MPEG encoding, we can decode each video frame directly for I frames or
from unlimited packet retransmission but also can reduce the FEC redundancy to conserve network bandwidth by predicting the effective packet loss rate. Meanwhile, when the end-to-end delay requirement can be met, the AHECM will only retransmit the necessary number of redundant FEC packets to receiver in comparison with legacy HARQ mechanisms. The AHECM uses an UEP to decide the appropriate FEC redundancy for different video frame types in order to produce the optimal amount of FEC redundancy. The AHECM uses the Markov model to approach the burst bit error condition over wireless net-works. The AHECM is evaluated by several metrics such as the effective packet loss rate, the ERE, the DFR, and PSNR to verify the efficiency in delivering video streaming over wireless networks. In the experiments evaluated with various metrics, the AHECM has outperformed the latest two HARQ mechanisms, i.e., the NASTE mechanism and the cross-layer error control mechanism. Furthermore, the AHECM has highlighted itself and impressed users with the visible high video quality. When delivering video streaming over wireless networks has become a trend today, the AHECM convinces us of its remarkable contri-butions and promising potential for being a multimedia transportation solution in the future.
References
1. Feher, G., Olah, I.: Enhancing wireless video streaming using lightweight approximate authentication. Multimed. Syst. 14(3), 167–177 (2008)
2. Tsai, M., Chilamkurti, N., Park, J., Shieh, C.: Multi-path trans-mission control scheme combining bandwidth aggregation and packet scheduling for real-time streaming in multi-path envi-ronment. IET Commun. 4(8), 937–945 (2010)
3. Han, R., Messerschmitt, D.: A progressively reliable transport protocol for interactive wireless multimedia. Multimed. Syst. 7(2), 141–156 (1999)
4. Ding, J., Tseng, S., Huang, Y.: Packet permutation: a robust transmission technique for continuous media streaming over the Internet. Multimed. Tools Appl. 21(3), 281–305 (2003) 5. Nafaa, A., Taleb, T., Murphy, L.: Forward error correction
strategies for media streaming over wireless networks. IEEE Commun. Mag. 46(1), 72–79 (2008)
6. Argyriou, A.: Cross-layer error control for multimedia streaming in wireless/wireline packet networks. IEEE Trans. Multimed. 10(6), 1121–1127 (2008)
7. Nguyen, D., Tran, T., Nguyen, T., Bose, B.: Wireless broadcast using network coding. IEEE Trans. Veh. Technol. 58(2), 914–925 (2009)
8. Dan, G., Fodor, V., Karlsson, G.: Robust source-channel coding for real-time multimedia. Multimed. Syst. 13(5–6), 363–377 (2008) 9. Soltani, S., Radha, H.: PEEC: a channel-adaptive feedback-based
error control protocol for wireless MAC layer. IEEE J. Sel. Areas Commun. 26(8), 1376–1385 (2008)
10. Sarkar, J., Sengupta, S., Chatterjee, M., Ganguly, S.: Differential FEC and ARQ for radio link protocols. IEEE Trans. Comput. 55(11), 1458–1472 (2006)
11. Tsai, M., Shieh, S., Ke, C., Deng, D.: A novel sub-packet forward error correction mechanism for video streaming over wireless networks. Multimed. Tools Appl. 47(1), 49–69 (2010)
12. Tsai, M., Shieh, C., Hwang, W., Deng, D.: An adaptive multi-hop FEC protection scheme for enhancing the QoS of video streaming transmission over wireless mesh networks. Int. J. Commun. Syst. 22(10), 1297–1318 (2009)
13. Moid, A., Fapojuwo, A.: Three-dimensional absorbing markov chain model for video streaming over IEEE 802.11 wireless networks. IEEE Trans. Consumer Electron. 54(4), 1672–1680 (2008)
14. Le, L., Hossain, E., Zorzi, M.: Queueing analysis for GBN and SR ARQ protocols under dynamic radio link adaptation with non-zero feedback delay. IEEE Trans. Wirel. Commun. 6(9), 3418–3428 (2007)
15. Boucheron, S., Salamatian, M.: About priority encoding trans-mission. IEEE Trans. Inf. Theory 46(2), 699–705 (2000) 16. Tsai, M., Chilamkurti, N., Zeadally, S., Vinel, A.: Concurrent
multipath transmission combining forward error correction and path interleaving for video streaming in wireless networks. Comput. Commun. doi:10.1016/j.comcom.2010.02.001
17. Tsai, M., Chilamkurti, N., Shieh, C., Vinel, A.: MAC-level forward error correction mechanism for minimum error recovery overhead and retransmission. Math. Comput. Model. doi:10.1016/ j.mcm.2010.05.019
18. Moid, A., Fapojuwo, A.: Heuristics for jointly optimizing FEC and ARQ for video streaming over IEEE 802.11 WLAN. IEEE Wireless Communications and Networking Conference, 2141–2146 (2008)
19. Subramanian, V., Kalyanaraman, S., Ramakrishnan, K.: Hybrid packet FEC and retransmission-based erasure recovery mecha-nisms for lossy networks: analysis and design. IEEE Int. Conf. Commun. Syst. Softw. Middlew. 1–8 (2007)
20. Kotuliakova, K., Polec, J.: Analysis of HARQ schemes using Reed-Solomon codes. IEEE Int. Conf. Syst. Signals Image Pro-cess. 323–326 (2008)
21. Tan, G., Herfet, T.: On the architecture of erasure error recovery under strict delay constraints. IEEE Eur. Wirel. Conf. 1–7 (2008) 22. Moon, S., Kim, J.: Network-adaptive selection of transport error control (NASTE) for video streaming over WLAN. IEEE Trans. Consumer Electron. 53(4), 1440–1448 (2007)
23. Won, Y., Ahn, S.: GOP ARIMA: modeling the nonstationarity of VBR processes. Multimed. Syst. 10(5), 359–378 (2005) 24. Krunz, M., Apostolopoulos, G.: Efficient support for interactive
scanning operation in MPEG-based video-on-demand systems. Multimed. Syst. 8(1), 20–36 (2000)
25. Shi, Y., Wu, C., Du, J.: A novel unequal loss protection approach for scalable video streaming over wireless networks. IEEE Trans. Consumer Electron. 53(2), 363–368 (2007)
26. Qu, Q., Pei, Y., Modestino, J.: An adaptive motion-based unequal error protection approach for real-time video transport over wireless IP networks. IEEE Trans. Multimed. 8(5), 1033–1044 (2006)
27. Gandikota, V., Tamma, B., Murthy, C.: Adaptive FEC-based packet loss resilience scheme for supporting voice communica-tion over ad hoc wireless networks. IEEE Trans. Mobile Comput. 7(10), 1184–1199 (2008)
28. Tsai, M., Huang, T., Shieh, C., Chu, K.: Dynamic combination of byte level and sub-packet level FEC in HARQ mechanism to reduce error recovery overhead on video streaming over wireless networks. Comput. Netw. doi:10.1016/j.comnet.2010.06.003
